Looking Ahead: The Impact of ChatGPT on the Marketing Industry

This article was originally published by GroupM China

With the ongoing invention of new words in the global advertising and marketing industry, marketers can often feel like they've been caught off guard or missed out on opportunities. For instance, "data clean rooms" in 2021, "metaverse" in 2022, and now “Generative AI” at the beginning of 2023. 

For each wave of hype, we attempt to evaluate the long-term impact it may have on the industry at its core. However, Generative AI is different because its long-term impact is about "existence," and related discussions often revolve around its immense creative potential or the possibility of replacing humans.

Last year, GroupM Global published "The Next 10: Artificial Intelligence" report, predicting that by 2032, 90% of media will use artificial intelligence in some way, and generative AI and machine learning algorithms would be the most critical artificial intelligence technologies as they can provide all parties in the media ecosystem with new insights and value sources.  

Regarding the recent hype surrounding generative AI - ChatGPT, GroupM Global's innovation department analyzed its potential impact on media and marketing in an internal opinion reference:

Search

The emergence of ChatGPT has sparked discussions about the significant changes that may occur in the future of search, user experience, and commercialization models. Mainstream search engines have already begun testing ChatGPT-like conversational AI technologies. Furthermore, advancements in this field may contribute to the further development of audio search.

Media

Artificial intelligence will continue to be used for recommendation engines and news information flow optimization, but more media platforms will begin testing the use of artificial intelligence to create content.

Creative

Generative AI marks the first time that computers have demonstrated "creative" ability, but there remains a gap between expectation and reality. The production aspect will be further disrupted. The democratization of creative tools on social media platforms has brought about many changes in content production. Giants such as Apple TV+ that have created their own content organizations have also promoted this disruptive change.

Generative AI can play a more significant role in the "production" process, but its "creative" ability is limited, especially when it comes to generating jokes, TV series, or art. There may be room for more improvements in subsequent application and platform-level developments.Media Planning and Activation

We need to approach individual data with caution. Integrated data allows us to generate data sets with "synthetic individuals." Although synthetic individuals cannot represent a real person in the real world, they can still help us understand the world, support insight generation, planning, and activation. By combining integrated data with other machine learning techniques, we can simulate the entire chain of consumer behavior to understand market trends, provide a basis for investment decisions for each brand, and support cross-channel planning.

In fact, GroupM and WPP have made significant investments in generative AI. In 2021, the data product and technology company Choreograph, launched by GroupM, has a team of data scientists in various machine learning and artificial intelligence fields, covering neural networks/GANs, agent-based modeling, and leveraging generative AI to drive creative insights. The same year, WPP acquired the company Satalia, which operates in all areas of artificial intelligence. WPP is also committed to promoting the "responsible and creative" use of generative AI.

To dissect the impact of ChatGPT on the marketing industry, we invited Head of GroupM China's Choreograph Vinny Ge, General Manager of GroupM Knowledge Zod Fang, GroupM China's Director of Brand Safety Jacob Zhou, Mindshare Data Analysis Director Winnie Wu, and EssenceMediacom's Director of Communication Strategy April Liu to share their takes on the subject:

Will ChatGPT also have an impact on the tag-based personalized recommendation model, and allow us to further understand the psychology and behavior of the target audience as well as recommend products with higher accuracy? 

Winnie：The search for a more accurate target audience is an eternal topic of marketing. The development of Internet technology in the last decade has greatly promoted the development of the ad industry and improved the efficiency and targeting ability of ad delivery. I believe that ChatGPT's excellent language comprehension ability can help improve the user profile and targeting ability of different media platforms in the future. I look forward to seeing ChatGPT or related AI technology being applied and integrated into more platforms and applications.

However, efficiency and advertising costs are also important factors to consider.

What functions can it deliver in brands' private domain operations? How will it change the existing model?

Winnie：Compared to the fixed and standard answers of current intelligent customer service chatbots, the advantage of ChatGPT is that it can understand questions in different contexts and provide more targeted and personalized responses. Essentially, ChatGPT has higher flexibility and adaptability. Therefore, ChatGPT can replace some of the work of manual customer service.

In the mid and short term, ChatGPT can serve as an effective supplement to the intelligent customer service system, providing users with faster and more practical support and services.

How will ChatGPT affect communication tools such as search engines? Will the precise Q&A mode of search services lead to the premature end of the search ad era?

Winnie：Although the combination of AI chatbots and search engines is indeed a trend, I think it is too early to say whether ChatGPT will end search engines or search advertising.

In addition, before the advent of ChatGPT, the search behavior of Chinese users had undergone significant changes. User search scenarios are getting more fragmented, and search channels more abundant. According to surveys, WeChat, Douyin, and RED search have become the primary search channels for many. Therefore, today's search ads not only refers to search engine ads, but also includes integrated search ad on multiple platforms. The rich content and forms of search ads on these platforms are currently beyond the capabilities of ChatGPT.

Jacob：The traditional search engine model provides many possible "answers" for each search query and allows users to decide the credibility of the information. When asking ChatGPT a question, the user may get a specific answer. This information is only valuable when the audience feels they can trust the answer, and gaining audience trust is an important issue that large language models face.

The ecosystem of the search market is always changing. The combination of ChatGPT and traditional search engines, at least at this stage, is more of an evolution than a revolution. The precise Q&A mode of search services will not directly lead to the rapid reduction of search result links and change the traditional model of search result ranking in the short term. What's more, search engines are committed to providing users with valuable experiences so that they can then serve ads and direct users to the advertiser's target website to earn profits. This industry model has stimulated a certain degree of symbiotic relationships between search engines and websites. Therefore, it is too early to say that the era of search advertising is about to end.

Will the content produced by AIGC surpass that of humans?

Vinny：ChatGPT is part of Generative AI, and language models are the most complex within this AI field. Other Generative AI are derived applications of image and audio-visual processing technologies.

Language itself is the expression of human thoughts and emotions, based on the accumulation of neural system information perception and feedback, and life experiences. This is something that language models do not possess. Although ChatGPT is a complex model trained on large datasets, the dimensions of information input into the model are still limited and one-sided. The model has a security layer that artificially sets boundaries and rules, imbuing it with human ethical and moral values. I believe that language models are currently still in the stage of trying to accurately comprehend human language, and do not have thinking capabilities. Therefore, the content generated or created by them has traces of information splicing. Hence, I do not think that the content produced by current Generative AI surpasses that of outstanding human works.

However, it is undeniable that AI Generated Content (AIGC) is starting to get close to human levels, which is a huge advancement in itself. Currently, AI can replace human labor to quickly produce lookalike content, significantly reducing the previously expensive costs of creation and creativity, and spawning new market spaces, even giving artists new inspiration. The bottleneck in current applications is the colossal computational resource costs of AI models and their dependence on data, but these issues can be quickly resolved.

In the marketing industry, content creativity has long dominated the field. Will ChatGPT replace some people's work functions? What does the most ideal model of work look like?

April: Based on my current experience with ChatGPT, I'd like to think about the relationship between ChatGPT and the marketing industry from two perspectives: "timeliness" and "professionalism". ChatGPT has excellent and efficient text processing capabilities, which will greatly improve our work efficiency in analyzing and summarizing past text and data. However, because ChatGPT is a chatbot developed by and based on a language learning model, it still has some limitations in handling and analyzing real-time events and hot new trends.

When the promoted text cannot be distinguished as human or machine-made, will this lead to the proliferation of "water armies"? Is this good or bad for brand marketing?  

Jacob：AI technology capable of making realistic content lowers the threshold for content production, expedites the process, and may promote the large-scale and covert use of "water armies". However, the root cause of the "water army" as a gray area industry lies in the driving force of the profit chain behind it, rather than the tools and technologies that assist the "water army", so AI technology is not a direct factor in "water army's” proliferation.

AI is at present the most advanced field in text generation. For brand marketing, this is a double-edged sword. On the one hand, generative AI can lend a quick and powerful hand in copywriting, brainstorming, and customer service conversations, keeping the marginal cost of content production and knowledge work very low, thus accelerating the production and distribution of creative content and empowering brand marketing. However, it should be noted that AI cannot replicate human biological attributes such as emotions and mental psychology. Technology is changing, but human nature remains the same. Texts that touch consumers' hearts, such as those with a "sincere" tone, still require human wisdom and guidance. The creation of AI is based on the induction, summary, deduction, and innovation of knowledge categories that have already been produced by thousands of humans. Compared with human creation, AI's high productivity may replace a large amount of basic creative work. Therefore, for brand marketing, it is necessary to understand the boundaries of AI's capabilities and make good use of AI, rather than being limited and constrained by AI.

On the other hand, large language models like ChatGPT now face issues such as inaccurate information production, bias, privacy and plagiarism, which means the text it generates may be false or inappropriate. This requires marketers to be on top of content regulation so as to avoid brand security risks in a semantic sense, reputational damage, profit loss, and even legal risks.

What challenges/obstacles will ChatGPT encounter in the realm of marketing?

Zod：Judging from the present, ChatGPT does have obvious advantages in efficiency compared to the interactive AI on the market, and such advantages are currently in the form of "toys" that everyone talks about, making it a high-traffic area.

However, when it comes to marketing, two questions need to be considered: What problems can it solve or help solve? Traffic, or sales problems? Or does it play a role in enhancing efficiency in the process of problem-solving (such as eCommerce website robo customer service)?  This issue ultimately comes down to a cost-benefit analysis. From current information, it is known that the cost of training a large ChatGPT model is in the range of millions of dollars, which is very expensive, and whether it can be economically efficient will ultimately determine how far ChatGPT can go.Secondly, ChatGPT relies on immense computing power, and behind that computing power is chips. When computing power develops on a large scale, ChatGPT may be able to achieve more growth, because continuing to develop iterations will require more parameters to be incorporated into the model.

A more important question on a different dimension is how can we regulate such a content production mode/mechanism? If ChatGPT wins people's trust for its accuracy, how will regulatory agencies regulate the content it produces? Especially when some content involves political/national security issues. In fact, there have been online requests for ChatGPT to evaluate different political figures, and there are issues with the model's biased answers. Content production mechanisms that are prone to affecting people's thinking will attract strong regulation, which is a problem that needs to be addressed.

To learn more about Mindshare China, scan the QR code and follow our official WeChat account.